Castellan
SIGNOZ OPENTELEMETRY NATIVE
Every agent decision,
governed in real time.
Castellan is the autonomous trust layer for AI. Intercept tool loops, token runaway spikes, and schema hallucinations in milliseconds before cost becomes failure.
Autonomous Intervention in 120ms
Click through the execution lifecycle to inspect how Castellan halts runaway agents.
STEP 01 // TIMESTAMP: 00:00.000
Agent Session Initiated
Code Reviewer Agent commences pull request audit. OpenTelemetry spans recorded.
GOVERNANCE_STATUS: ACTIVE
INTERVENTION_LATENCY: 120ms
PREVENTED_DAMAGE: $50.00
The Three Engines
Detect, decide, and explain every autonomous action.
01 / ANOMALY DETECTION
Sentinel Engine
Consumes OpenTelemetry spans in real-time. Detects tool loops, cost spikes, schema hallucinations, and behavioral drift.
02 / GOVERNANCE ENGINE
Tribunal Engine
Evaluates anomalies against YAML policies. Automatically executes Kill, Throttle, Rollback, or Escalate actions.
03 / REPLAY & AUDIT
Chronicle Engine
Reconstructs visual decision replays with plain-English root causes and exports downloadable executive PDF reports.